AI visibility is not a result you achieve and keep. Models are replaced, retrained and revised; sources shift; competitors become active. What is true today can look different in two months — without anything having changed on your side.
That is why the path does not end at execution but at re-measurement, and starts again afterwards. The same questions, the same channels, the same yardstick, at fixed moments. That is how you separate movement from noise. And by laying your own interventions and control questions alongside it, you can estimate where that movement came from. Model and product changes we record separately, so a jump in the figures is not quietly put down to your account.
That last point matters more than it sounds. When a model suddenly shows fewer sources, in a dashboard it looks like a collapse — while your presence in the answer text barely moved. Read citations apart from mentions and you draw the wrong conclusion, adjusting for something that did not happen.
The recurring measurement runs on our own methodology: the same question sets, the same channels, the same yardstick as the baseline. You get the numbers, but above all the reading: what shifted, what it means, and what we do with it in the next round.